cars.applications.point_cloud_outlier_removal.outlier_removal_tools

This module contains functions used in outlier removal

Module Contents

Functions

small_component_filtering(→ Tuple[pandas.DataFrame, ...)	Filter point cloud to remove small clusters of points
detect_small_components(→ List[int])	Determine the indexes of the points of cloud_xyz to filter.
statistical_outlier_filtering(...)	Filter point cloud to remove statistical outliers
detect_statistical_outliers(→ List[int])	Determine the indexes of the points of cloud_xyz to filter.
epipolar_small_components(cloud, epsg[, ...])	Filter outliers using the small components method in epipolar geometry
epipolar_statistical_filtering(epipolar_ds, epsg[, k, ...])	Filter outliers using the statistical method in epipolar geometry

cars.applications.point_cloud_outlier_removal.outlier_removal_tools.small_component_filtering(cloud: pandas.DataFrame, connection_val: float, nb_pts_threshold: int, clusters_distance_threshold: float = None, filtered_elt_pos: bool = False) → Tuple[pandas.DataFrame, Union[None, pandas.DataFrame]]

Filter point cloud to remove small clusters of points (see the detect_small_components function).

Parameters

• cloud – combined cloud as returned by the create_combined_cloud function

• connection_val – distance to use to consider that two points are connected

• nb_pts_threshold – number of points to use to identify small clusters to filter

• clusters_distance_threshold – distance to use to consider if two points clusters are far from each other or not (set to None to deactivate this level of filtering)

• filtered_elt_pos – if filtered_elt_pos is set to True, the removed points positions in their original epipolar images are returned, otherwise it is set to None

Returns

Tuple made of the filtered cloud and the removed elements positions in their epipolar images

cars.applications.point_cloud_outlier_removal.outlier_removal_tools.detect_small_components(cloud_xyz: numpy.ndarray, connection_val: float, nb_pts_threshold: int, clusters_distance_threshold: float = None) → List[int]

Determine the indexes of the points of cloud_xyz to filter. The clusters are made of ‘connected’ points (2 connected points have a distance smaller than connection_val)

The removed clusters are composed of less than nb_pts_threshold points and are also far from other clusters (points are further than clusters_distance_threshold).

If clusters_distance_threshold is set to None, all the clusters that are composed of less than nb_pts_threshold points are filtered.

Parameters

• cloud_xyz – points kdTree

• connection_val – distance to use to consider that two points are connected

• nb_pts_threshold – number of points to use to identify small clusters to filter

• clusters_distance_threshold – distance to use to consider if two points clusters are far from each other or not (set to None to deactivate this level of filtering)

Returns

list of the points to filter indexes

cars.applications.point_cloud_outlier_removal.outlier_removal_tools.statistical_outlier_filtering(cloud: pandas.DataFrame, k: int, dev_factor: float, use_median: bool = False, filtered_elt_pos: bool = False) → Tuple[pandas.DataFrame, Union[None, pandas.DataFrame]]

Filter point cloud to remove statistical outliers (see the detect_statistical_outliers function).

Parameters

• cloud – combined cloud as returned by the create_combined_cloud function

• k – number of neighbors

• dev_factor – multiplication factor of deviation used to compute the distance threshold

• use_median – choice of statistical measure used to filter

• filtered_elt_pos – if filtered_elt_pos is set to True, the removed points positions in their original epipolar images are returned, otherwise it is set to None

Returns

Tuple made of the filtered cloud and the removed elements positions in their epipolar images

cars.applications.point_cloud_outlier_removal.outlier_removal_tools.detect_statistical_outliers(cloud_xyz: numpy.ndarray, k: int, dev_factor: float, use_median: bool) → List[int]

Determine the indexes of the points of cloud_xyz to filter. The removed points have mean distances with their k nearest neighbors that are greater than a distance threshold (dist_thresh).

This threshold is computed from the mean (or median) and standard deviation (or interquartile range) of all the points mean distances with their k nearest neighbors:

(1) dist_thresh = mean_distances + dev_factor * std_distances or (2) dist_thresh = median_distances + dev_factor * iqr_distances

Parameters

• cloud_xyz – points kdTree

• k – number of neighbors

• dev_factor – multiplication factor of deviation used to compute the distance threshold

• use_median – if True formula (2) is used for threshold, else formula (1)

Returns

list of the points to filter indexes

cars.applications.point_cloud_outlier_removal.outlier_removal_tools.epipolar_small_components(cloud, epsg, min_cluster_size=15, radius=1.0, half_window_size=5, clusters_distance_threshold=np.nan)

Filter outliers using the small components method in epipolar geometry

Parameters

• epipolar_ds (xr.Dataset) – epipolar dataset to filter

• epsg (int) – epsg code of the CRS used to compute distances

• statistical_k (int) – k

• std_dev_factor (float) – std factor

• half_window_size (int) – use median and quartile instead of mean and std

• use_median (bool) – use median and quartile instead of mean and std

Returns

filtered dataset

Return type

xr.Dataset

cars.applications.point_cloud_outlier_removal.outlier_removal_tools.epipolar_statistical_filtering(epipolar_ds, epsg, k=15, dev_factor=1.0, half_window_size=5, use_median=False)

Filter outliers using the statistical method in epipolar geometry

Parameters

• epipolar_ds (xr.Dataset) – epipolar dataset to filter

• epsg (int) – epsg code of the CRS used to compute distances

• statistical_k (int) – k

• std_dev_factor (float) – std factor

• half_window_size (int) – use median and quartile instead of mean and std

• use_median (bool) – use median and quartile instead of mean and std

Returns

filtered dataset

Return type

xr.Dataset